A Preliminary RSWKS Framework for Multiple Chang Points (MCPs) Detection on Pathological Signal Analysis in Partial Epilepsy

In order to better interpret and early detect the heart abnormalities, in this paper a preliminary RSWKS framework describes the implementation of multiple change-points (MCPs) detection on pathological ECG signal analyses of post-ictal heart rate oscillations in partial epilepsy. In our random slide window (RSW) method, the slide window size is randomly selected from an optional set predefined by practical data features and experimental knowledge, then the resultant MCPs are obtained by means of KS statistics from each slide window. By comparing with the fixed slide window (FSW) approach, the proposed RSWKS is evaluated by using synthetic time series, as well as ECG samples of postictal heart rate oscillations in partial epilepsy. Then, the abnormal patterns are recognized by analyzing data features among the resultant multiple change points. Moreover, the severity of partial epilepsy is roughly distinguished in terms of the number, strength and period time of ECG oscillations in the stage of partial epilepsy sudden attack. The purpose of our RSWKS is to provide a novel platform for abnormal pattern recognition through MCPs detection quickly and efficiently.